Most of the recent progresses on visual question answering are based recurrent neural networks (RNNs) with attention. Despite success, these models often timeconsuming and having difficulties in modeling long range dependencies due to sequential nature RNNs. We propose a new architecture, Positional Self-Attention Coattention (PSAC), which does not require RNNs for video answering. Specifically, inspired by success self-attention machine translation task, we calculate response at each...

Recent progress has been made in using attention based encoder-decoder framework for image and video captioning. Most existing decoders apply the mechanism to every generated word including both visual words (e.g., “gun” “shooting”) non-visual “the”, “a”). However, these can be easily predicted natural language model without considering signals or attention. Imposing on could mislead decrease overall performance of Furthermore, hierarchy LSTMs enables more complex representation data,...

Existing video hash functions are built on three isolated stages: frame pooling, relaxed learning, and binarization, which have not adequately explored the temporal order of frames in a joint binary optimization model, resulting severe information loss. In this paper, we propose novel unsupervised hashing framework dubbed self-supervised (SSVH), is able to capture nature videos an end-to-end learning fashion. We specifically address two central problems: 1) how design encoder-decoder...

Video visual question answering (V-VQA) remains challenging at the intersection of vision and language, where it requires joint comprehension video natural language question. Image-Question co-attention mechanism, which aims generating a spatial map highlighting image regions relevant to vice versa, has obtained impressive results. Despite success, simply applying results in unsatisfactory performance due complexity temporal nature videos. In this paper, we proposed novel architecture,...

Visual question answering (VQA) that involves understanding an image and paired questions develops very quickly with the boost of deep learning in relevant research fields, such as natural language processing computer vision. Existing works highly rely on knowledge data set. However, some require more professional cues other than set to answer correctly. To address issue, we propose a novel framework named knowledge-based augmentation network (KAN) for VQA. We introduce object-related...

Fully mining visual cues to aid in content understanding is crucial for video captioning. However, most state-of-the-art captioning methods are limited generating captions purely based on straightforward information while ignoring the scenario and context information. To fill gap, we propose a novel, simple but effective scenario-aware recurrent transformer (SART) model execute Our contains “scenario understanding” module obtain global perspective across multiple frames, providing specific...

Generating consecutive descriptions for videos, that is, video captioning, requires taking full advantage of visual representation along with the generation process. Existing captioning methods focus on an exploration spatial-temporal representations and their relationships to produce inferences. However, such only exploit superficial association contained in a itself without considering intrinsic commonsense knowledge exists dataset, which may hinder capabilities cognitive reason accurate...

ChatGPT has been emerging as a novel information source, and it is likely that the public might seek from while taking protective actions when facing climate hazards such floods hurricanes. The objective of this study to evaluate accuracy completeness responses generated by individuals about aspects actions. survey analysis results indicated that: (1) emergency managers considered provided accurate complete great extent; (2) was statistically verified in evaluations accurate, but lacked...

Loitering detection can help recognize vulnerable people needing attention and potential suspects harmful to public security. The existing loitering methods used time or target trajectories as assessment criteria, only handled some simple circumstances because of complex track. To solve these problems, this paper proposes a method based on pedestrian activity area classification. first gave definition from new perspective using the size area. behaviors were divided into three categories....

Recent progress has been made in using attention based encoder-decoder framework for image and video captioning. Most existing decoders apply the mechanism to every generated word including both visual words (e.g., "gun" "shooting") non-visual (e.g. "the", "a"). However, these can be easily predicted natural language model without considering signals or attention. Imposing on could mislead decrease overall performance of Furthermore, hierarchy LSTMs enables more complex representation data,...

This paper presents a new approach for constructing decision trees based on variable precision rough set model. The presented is aimed at handling uncertain information during the process of inducing and generalizes to tree construction by allowing some extent misclassification when classifying objects. In paper, weighted mean are introduced. algorithm effectively overcomes influence noise data in structuring tree, reduces complexity strengthens its extensive ability.

Wireless sensor networks have been widely applied to various application domains such as environmental monitoring and surveillance. Because of reliance on the open transmission media, a network may suffer from radio jamming attacks, which is easy launch but difficult defend. Attacked by signals, experience corrupted packets low throughput. A number defense techniques proposed. However, each technique suitable for only limited range conditions. This paper proposes an adaptive approach...

It is a common fact that data (features, characteristics or variables) are collected at different sampling frequencies in some fields such as economic and industry. The existing methods usually either ignore the difference from hardly take notice of inherent temporal mixed frequency data. authors propose an innovative dual attention-based neural network for (MID-DualAtt), order to utilize select input reasonably without losing information. According authors' knowledge, this first study use...

Wireless sensor networks have been widely applied to various application domains such as environmental monitoring and surveillance. Because of reliance on the open transmission media, a network may suffer from radio jamming attacks, which are easy launch but difficult defend. Attacked by signals, experience corrupted packets low throughput. A number defense techniques proposed. However, each technique is suitable for only limited range conditions. This paper proposes an adaptive approach...

Aiming at the problem that counter propagation network (CPN) can not make good use of nerve cells, a rough CP neural model based on set is proposed. It changes strategy winning as king, and decides output values to Rough Member Function, which expresses level an element subordinated set. Experiments show approach solve some problems in other Networks, for example, sample size quality. They would directly influence accuracy. While reducing training time, prediction precision be greatly improved.

A new thresholding algorithm and its multi-threshold extension are presented to improve the performance of image segmentation. The probabilistic rough set is used describe objects backgrounds in image. optimal threshold given through a minimum boundary criterion. membership function defined light requirements thresholding.

Visual dialog aims to answer several consecutive questions based on image and history. Most works resolve all with ambiguous references (e.g., "she") by history, which generates redundant information gets in-accurate results. Also, they regard this task as a classification task, ignores the diversity of response answers results in poor generalization capability. To tackle these problems, we propose novel Context Gating Multi-level Ranking Learning (CGMRL). Specifically, proposed context...

Based on the rough set theory, a counter propagation neural network algorithm for edge detection is presented in this paper. Firstly, definition of membership function, which used to modify weigh values nomal network, proposed after introducing set. Experiments show that approach has achieved good results improving accuracy detection. And can also overcome effectively problem simply cluster network.